signal · 01
Vibration · triaxial
IEPE accelerometer · SKF · Hansford
Spectrum signature catches bearing wear, imbalance, mis-alignment and looseness. ISO 10816 baselines.
6 sensor signal types
What does the system actually measure on your machines?
Six signal types cover 90% of failure modes on rotating equipment. Daxonet specifies the exact stack per machine during the free audit · you only pay for sensors that earn their keep.
signal · 02
Temperature · RTD
Banner Engineering · Honeywell
Bearing surface, motor casing, oil sump. Trend deltas catch lubrication failure and cooling problems.
signal · 03
Motor current · CT
Schaeffler · Banner · stock CTs
Clamp-on current transformers detect rotor bar damage, broken windings, voltage imbalance.
signal · 04
Acoustic · ultrasonic
UE Systems · Hansford
35–60 kHz airborne emission. Catches lubrication starvation, valve leaks, electrical arcing very early.
signal · 05
Oil quality
Trico · Tan Delta
In-line viscosity, water content, particle counter. Prevents lubrication-induced bearing failures.
signal · 06
Pressure · differential
Honeywell · WIKA
Filter clog, pump cavitation, compressor degradation. Differential between inlet and outlet.
6 standard failure modes
Which failure modes does the model actually detect?
Six classes covered out of the box on rotating equipment. Custom modes for your specific assets are added in 2 to 4 week training cycles after enough labelled history is collected.
mode · 01
Bearing wear · pitting
14–30 d lead
Vibration spectrum signature in BPFI/BPFO frequencies. Catches inner-race and outer-race defects very early.
detection confidence
88%
mode · 02
Motor electrical imbalance
10–21 d lead
Current signature analysis (MCSA). Detects broken rotor bars, stator winding damage, voltage imbalance.
detection confidence
74%
mode · 03
Shaft mis-alignment
7–21 d lead
Vibration at 2× running speed plus axial vibration. Catches angular and parallel mis-alignment.
detection confidence
70%
mode · 04
Lubrication breakdown
7–14 d lead
Acoustic emission and oil sensor delta. Catches bearing starvation before metal-on-metal contact.
detection confidence
60%
mode · 05
Mechanical looseness
14–30 d lead
Multiple harmonics in vibration. Foundation bolts, bearing housing, coupling guard. Easy to schedule.
detection confidence
80%
mode · 06
Pump cavitation
3–10 d lead
Random high-frequency vibration plus acoustic emission. Catches NPSH problems before impeller damage.
detection confidence
42%
End-to-end pipeline
How does an alert actually get to your maintenance team?
Five stages from sensor signal to scheduled intervention. Edge inference means the line keeps running even if internet is down.
stage · 01
continuous
Sense
Sensors stream signal at 1 to 25 kHz to the edge box next to the asset. Triaxial vibration, temperature, current, acoustic.
stage · 02
~50 ms
Stream
Edge box ingests, filters and feature-engineers the signal. FFT for vibration, MCSA for current. Buffered locally.
stage · 03
~200 ms
Predict
Trained anomaly model scores the live features against the asset baseline. Outputs failure mode, confidence and lead time.
stage · 04
~1 s
Alert
Watch / Warning / Critical alert routed to maintenance team via Teams, WhatsApp or email. Includes failure mode and recommended action.
stage · 05
CMMS · ERP
Schedule
Work order auto-created in your CMMS with asset ID, failure mode, recommended intervention, parts and lead time. Logged for audit.
Production savings · typical
What does this actually save us per year?
Typical year-one numbers from Daxonet plant deployments. Actual savings depend on current breakdown frequency, downtime cost-per-hour and asset criticality.
30–50%
unplanned downtime
reduced vs reactive baseline
18–24 mo
machine life extended
critical rotating assets
25–40%
maintenance overtime
cut from emergency calls
9 mo
typical payback
one or two avoided breakdowns
6 manufacturing verticals
Where does this deliver fastest payback in Malaysia?
Six verticals where Daxonet has shipped predictive-maintenance pilots. Each has a starter asset list, sensor recipe and CMMS integration playbook.
Pumps
Compressors
Mixers
Conveyors
CNC spindles
SMT lines
HVAC
Air compressors
Stamping presses
Robots
Conveyors
Welders
Hydraulic pumps
Fans
Crane motors
Compressors
Centrifugal pumps
Agitators
Heat exchangers
Compressors
Cooling towers
Gensets
Transformers
Pumps
Architecture · integration
How does it integrate with our CMMS, MES and ERP?
Native integration with Microsoft Azure IoT, arc.ops MES, D365 and standard CMMS platforms. Edge-first means production keeps running even if internet is down.
node · 01
Sensor · field
SKF · Banner · Hansford · Honeywell
node · 02
Edge AI
Jetson · NUC · industrial PC
node · 03
Azure IoT · cloud
Storage · retraining · dashboards
node · 04
CMMS · work orders
Maximo · eMaint · UpKeep · Fiix
node · 05
D365 · ERP
Asset cost · OEE · KPIs
Edge-first by design. Anomaly inference runs on the edge box on the factory floor, so machine monitoring keeps working even if internet is down. Cloud is used only for retraining, dashboards and historical analytics · never for real-time alerts.
6-week pilot · live
How long until alerts start hitting our maintenance team?
Six weeks from contract signature to live pilot on three critical machines. Fixed-price after the free machine-health audit.
01
Week 1
Audit
Asset criticality scoring. Sensor specification per machine. Existing PLC and SCADA tag inventory. Free deliverable.
02
Week 2
Install
Sensor mounting, wiring and edge box rack-up next to the line. Power, network, signal validation by Daxonet engineers.
03
Week 3
Baseline
Continuous data collection on healthy machine. Model learns what good looks like for your specific assets.
04
Week 4
Tune
Anomaly model trained on the baseline. Threshold and confidence calibration per machine and per failure mode.
05
Week 5
Integrate
CMMS work-order creation. Alert routing to Teams, WhatsApp, email. Dashboard handed to maintenance planner.
06
Week 6
Live
Supervised parallel run · model alerts but maintenance team also runs normal walkdown. Confidence builds, automation goes live.
FAQ · plant managers ask
Common questions before booking the free machine-health audit.
01Which machines are best suited for AI Predictive Maintenance?
Rotating equipment with measurable failure signatures gets the fastest payback: electric motors, pumps, fans, blowers, gearboxes, compressors, conveyor drives, and CNC spindles. Reciprocating equipment (piston pumps, compressors) and bearings on critical axes also benefit. The system works on assets from 0.5 kW up to 500 kW. The starting question Daxonet asks during the free audit is which 3 to 5 machines, if they fail unplanned, would cause the largest production loss · those become the pilot scope. Static equipment like vessels and pipework is a different problem (corrosion, leak detection) and needs a different sensor stack.
02What can the model actually predict · what failure modes does it cover?
Six standard failure modes are covered out of the box: bearing wear and pitting, motor electrical imbalance, shaft mis-alignment, lubrication breakdown, mechanical looseness, and cavitation in pumps. Each is detected with a different signal: vibration spectrum signatures for bearings and imbalance, current signature for motor faults, oil sensor data for lubrication breakdown, acoustic for cavitation, temperature deltas for cooling and bearing degradation. Custom failure modes specific to your equipment can be added in 2 to 4 week training cycles once enough labelled history is collected.
03How early does the model warn us before actual failure?
Lead time is 7 to 30 days for most rotating-equipment failure modes when sensors are installed correctly and baselines are properly captured. Bearing wear typically gives 14 to 30 days lead time. Motor electrical faults give 10 to 21 days. Lubrication breakdown gives 7 to 14 days. Cavitation gives 3 to 10 days. The system reports both an alert level (Watch, Warning, Critical) and a confidence score, so your maintenance planner can schedule the intervention into the next planned shutdown rather than reacting to an unplanned breakdown.
04What sensors do we need · do we need to retrofit our machines?
Most factories install a combination of triaxial vibration sensors (IEPE accelerometers), surface or RTD temperature sensors, motor current transformers (CTs) clamped on the motor cable, and optional ultrasonic acoustic sensors for valves and bearings. Daxonet works with industrial brands like SKF, Banner Engineering, Hansford, Honeywell and Schaeffler. Wireless sensors are used where cabling is impractical. Existing PLC and SCADA tags are also ingested where available, so you do not duplicate hardware.
05How long until the pilot is live on our first machines?
6 weeks from contract signature to live pilot on 3 critical machines. Week 1 is the asset audit and sensor specification. Week 2 is sensor installation and wiring. Week 3 is baseline data collection (the model learns what healthy looks like for your specific machines). Week 4 is anomaly model tuning and threshold calibration. Week 5 is integration with your CMMS and alert channels (Teams, WhatsApp, email). Week 6 is supervised parallel running where the model alerts but maintenance team also runs their normal walkdown, building trust before automation goes live.
06How does it integrate with our CMMS, ERP or MES?
Native integration with Microsoft Dynamics 365 Supply Chain Management asset management, Arcstone arc.ops MES, and standard CMMS platforms (IBM Maximo, eMaint, UpKeep, Fiix, plus simpler Excel-driven workflows). When the model flags a Warning or Critical alert, a work order is auto-created with the asset ID, predicted failure mode, recommended intervention and estimated lead time. The original signal trace is attached for the technician to verify. PLC, SCADA and OPC UA tag ingestion is also supported for plants that already have data historians.
07Is the data secure · what about our process recipes and IP?
Yes. Sensor data and the trained anomaly model stay on your edge hardware or your private Azure tenant, not on a shared cloud. Daxonet does not use your machine signatures to train models for other clients · this is contractually guaranteed. PDPA controls apply to any human imagery captured incidentally. Audit trail is enabled on every alert and intervention recommendation with 7-year retention by default. Daxonet's security team signs off the architecture before installation, and we provide the documentation your customer audit or compliance officer needs.
08What does AI Predictive Maintenance actually cost us?
Pricing is scoped to your size, modules and integrations. Daxonet quotes fixed-price after a short scoping call so there are no surprises. Most clients reach payback within the same project window.
Related on Daxonet